Use of AI in music...

GtrGeorge

Hi,
AI is here and the music industry is responding in various ways. I hope that ST4 is developed with consideration to the important pervasive changes we are all going thru.
To musicians who are affiliated with a Performing Rights Organizations (PROs), such as ASCAP,BMI and SOCAN, your registered creative work now falls under their policies. What does that mean? It means if you used AI to help create the music you may not get paid the same way or receive credit the same as in the past. You could loose money simply by using AI Mastering Software on your tracks. Even more if you used Chat GPT to create lyrics. Maybe you receive nothing if you used AI substantially. And no credit (attribution).
I can't forsee all of the changes AI will bring to the music industry. But I can say that I use ST out live to play my music in venues and if future ST versions use AI...I may have to leave ST Forever, because i would (potentially) be reducing my income.
We ARE NOT THERE YET. But I am concerned so I post this here ,where Peter will see it and give it some thought..and you, as well. Musicians who aren't involved with PROs, or use other PROs may have very different circumstances, so keep that in mind.
George Barry

PS: here is a link to an attorney discussing the PROs announcement. Not comprehensive but a primer on what the Big PROs said. https://www.youtube.com/watch?v=YEQtjzr6f78

AnthonyB

It's an interesting and complex topic, AI in music, and even more boradly, AI in any creative works (art, literature, etc.).
I also think that when it comes to certain plugins in particular (i.e. mastering plugins) that there is often some extremely loose use of the term AI (because if your product doesn't contain "AI" then it isn't as good and modern as others.. marketing.. sheesh..) which really should be the pre-cursor, ML (machine learning). Anyway, this topic in a broad sense could go down all sorts of rabbit holes...

..back to AI and ST. I have definitely thought about this. What could AI add to an app like ST to really add value, or improve some part of it in some way?
Based on what I see of ST3/4, ST isn't a tool that you create in, so we're not at risk of the Suno/Udio/etc. AI song creation type situation.
Audio processing? I would tend to think that ST, isn't trying to be an audio processor beyond a 'lighter touch' approach. So are you trying to use Izotope Ozone or similar 'heavy' mastering tools that claim to use some sort of AI to enhance your music? No. Would you even want to use some sort of AI processing on an iPad, in a live performance? I think the latency and inconsistency would be a killer unless some clever boffin manages to package a usable local LLM in an iOS app. I wouldn't rule that out.

The only applications that I've thought off that might be useful and add value to the user experience in ST tend to be off-line applications, e.g. Having an AI in the lyric editor to perform more complex search and replace operations (I currently do this outside of ST and import the results), or in Song List 'smart filtering', e.g. 'find me other songs like this one' that goes beyond 'similar tempo' or 'similar year value' and performs audio analysis, categorization, etc. That said, ST could implement most of that in some sort of metadata lookup from existing internet sources and store it with the song information...

Anyone, quite a tricky topic. I think if I was to pose what I think is a fairly neutral position of "what are the acceptable uses of AI in a product like Stage Traxx?" that could still be polarising to some people.

Vedeja

An AI usecase for ST that I dreamt up just now:

ST is listening to vocal inputs live. It does speech to text and an LLM analyzes the speech. The vocalist yells ”let’s break it down!”. ST begins modifying the mix by, step by step fading, muting, EQ:ing out tracks. As the song progresses and breaks down the band keeps saying things like ”solo” or ”c’mon”, further triggering the AI driven automixer to adjust the production.

I think it would be interesting, and would it fit the case stated by @GtrGeorge?

peter

Currently on my long term todo list is indeed the use of a custom trained AI subsystem for detecting the beatgrid, tempo and key of songs as that would open lots of useful stuff during playback like synchronized metronome, exact jumps at the end of a bar anywhere in the song, adhoc looping anywhere in the song, better transposition features.

But that‘s all I have on my own wishlist for now regarding AI.

Behind the scenes I already make heavy use of AI. Maybe 80% of the code in ST4 is already written by AI.